Abstract: An adult image filtering method based on multi-features is designed and implemented. The authors detect the skin and texture feature in image and extract some statistical features based on skin and texture features. The features are input into a decision tree and the image is classified. Experiments show that 90% images could be recognized accurately.